LETTERdoi:10.1038/nature13487

An optoelectronic framework enabled bylow-dimensional phase-change filmsPeiman Hosseini1, C. David Wright2 & Harish Bhaskaran1

The development of materials whose refractive index can be opticallytransformed as desired, such as chalcogenide-based phase-changematerials, has revolutionized the media and data storage industriesby providing inexpensive, high-speed, portable and reliable platformsable to store vast quantities of data. Phase-change materials switchbetween two solid states—amorphous and crystalline—in response toa stimulus, such as heat, with an associated change in the physical pro-perties of the material, including optical absorption, electrical con-ductance and Young’s modulus1–5. The initial applications of thesematerials(particularlythegermaniumantimonytelluriumalloyGe2Sb2Te5)exploited the reversible change in their optical properties in rewri-table optical data storage technologies6,7. More recently, the changein their electrical conductivity has also been extensively studied in thedevelopment of non-volatile phase-change memories4,5. Here we showthat by combining the optical and electronic property modulation ofsuch materials, display and data visualization applications that go beyonddata storage can be created. Using extremely thin phase-change materialsand transparent conductors, we demonstrate electrically inducedstable colour changes in both reflective and semi-transparent modes.Further, we show how a pixelated approach can be used in displayson both rigid and flexible films. This optoelectronic framework usinglow-dimensional phase-change materials has many likely applications,such as ultrafast, entirely solid-state displays with nanometre-scalepixels, semi-transparent ‘smart’ glasses, ‘smart’ contact lenses and arti-ficial retina devices.

Phase-change materials (PCMs) have been key to enabling the creationof a versatile platform able to store vast amounts of information on inex-pensive plastic substrates (DVDs and other optical media6,7) using opticalpulses. These technologies use light of specific wavelengths (for example, Blu-Ray at 405 nm, DVDs at 650 nm and CDs at 780 nm) to write and read data.The next generation of electronic, solid-state memories based on PCMscould replace the current leading storage technologies, namely Flash andmagnetic disks. PCMs1–5 have extraordinary properties such as extremescalability, fast switching speeds, and high switching endurance4,5. How-ever, although PCMs have been commercialized in both the electrical4

and optical domains8, there has been very little research into understand-ing how transforming the phase in one domain affects the properties inanother (that is, how switching electrically affects optical properties, andvice versa). This is because the likely applications emerging from suchnon-volatile, optoelectronic properties have so far been unclear. Herewe propose and demonstrate these optoelectronic properties in the con-text of novel display and data visualization applications.

Recently, thin-film perfect absorbers in the mid-infrared part of thespectrum have been demonstrated, where the optical properties of atemperature-transient PCM (vanadium dioxide, VO2) were tuned9. VO2,however, is stable in each of its phases only within certain temperatureranges, a property that makes it energy inefficient and impractical to im-plement in electrically controlled circuits at the nanoscale. In contrast, theprimary phases of Ge2Sb2Te5 (GST)—amorphous and crystalline—arestable for most applications10 and can be thermally, optically or elec-trically switched at ultrahigh speed11,12. Furthermore, GST is extremely

scalable13,14, easily integrated in commercial devices and has optical prop-erties that can be tuned in the visible spectrum of light15.

We describe a unique optoelectronic framework technology employingGST-based thin films and experimentally demonstrate its wide-rangingapplicability in various types of display technologies. We show how such asystem, when combined with a transparent electrode such as indium tinoxide (ITO), can be used as a reflective or a semi-transparent microdisplayand on both rigid and flexible substrates, which could usher in an era oftruly flexible electronic screens that can operate at the maximum pos-sible resolution allowed by the diffraction of visible light, both as back-lit displays and electronic readers. Current microdisplay technologies basedon liquid crystals, microelectromechanical systems and organic light-emitting diodes are attracting considerable attention because of a growinginterest in wearable technology16. Key requirements for such applicationsare high resolution, high speed and low power consumption, all of whichare met by the technology described here.

We first describe the reflective type of display using GST-based thinfilms. Figure 1a shows the cross-sectional schematic of our reflective devices;the GST is sandwiched between two ITO layers and deposited on topof a reflective surface such as platinum. The crystallization of the few-nanometres-thick GST layer induces a colour change in the entire film,visible when incident ‘white’ light is reflected back. To understand the re-lationship between the thickness of the ITO and GST layers and the overalloptical properties of the stack we systematically compute the reflectivityspectra of the stack while gradually increasing the thickness of each layer.Every calculation is made twice, one for each phase of the GST layer; achange in the refractive index of the GST layer (caused by a change inphase) modulates the reflective spectrum of the entire stack. The spectraare finally compared and the percentage change in reflectivity is plotted

asDR (%)~Rcrystal{Ramorphous

Ramorphous| 100

� �, with Rcrystal being the reflec-

tivity (at a specific wavelength) of the entire stack when the GST layer is inthe crystalline phase and Ramorphous its reflectivity when the GST layer is inthe amorphous phase. We use the above expression because the amorph-ous phase is the as-deposited phase in our experiments. These models (topgraphs in Fig. 1b) indicate that one key factor is the presence of a trans-parent conductor beneath the GST (with thickness t in Fig. 1a), which werefer to as the ITO spacer. This layer is crucial and contributes to the colourin our system.

To experimentally verify our reflectivity calculations, we sputter anumber of multi-layered films on silicon wafers with a 300-nm-thickthermally grown silicon dioxide (SiO2) layer; our GST layer is knownto be amorphous as sputtered (section S4 in the Supplementary Infor-mation). Figure 1d shows a photograph of a few sputtered films with similarstructure (10 nm ITO/7 nm amorphous GST/t nm ITO/100 nm Pt/SiO2);t is the only thickness that is varied. The colour of the films when t (theITO spacer thickness) is varied varies dramatically. To fully crystallize theGST, we heat each sample up to 220 uC for a few minutes on a hot plate.The observed variation in the colour of these samples when the GST iscrystallized is also shown in Fig. 1d; the colour variations show the

1Department of Materials, University of Oxford, Parks Road, Oxford OX1 3PH, UK. 2College of Engineering, Mathematics and Physical Sciences, University of Exeter, Harrison Building, North Park Road,Exeter EX4 4QF, UK.

2 0 6 | N A T U R E | V O L 5 1 1 | 1 0 J U L Y 2 0 1 4

Macmillan Publishers Limited. All rights reserved©2014

potential for the use of such films as individual pixels in display devices,each pixel being a different colour to achieve a standard colour displaydevice (and with switching in real devices achieved electrically; see below).

To quantify this colour change, we first model the system using atransfer matrix optical computational method17 to calculate propertiessuch as reflectance, transmittance and internal electrical field at visiblewavelengths (full details of the model are available in section S5 of theSupplementary Information). We then carry out reflectivity measure-ments on sputtered films to compare our model with experiments; weuse a spectrometer (Lambda 1050, Perkin Elmer US) to obtain the refle-ctivity spectra of the samples both before and after crystallization. Figure 1b

shows exceptionally good agreement between the simulated (top, left)and measured (bottom, left) reflectivity spectra for the samples shownin Fig. 1d, when the thin GST layer is in the amorphous phase. Similarreflectivity spectra are shown (Fig. 1b, right) for the same samples aftercrystallizing the GST layer, where again the agreement between experi-ment and theory is very good. These results show that an effective changein the reflectivity spectrum (that is, colour) of the stack can be inducedsimply by changing the phase of a chalcogenide layer a few nanometresthick.

The next aspect of our study focuses on the influence of the GST thick-ness. In conventional optical disks, the thickness of the GST layer is

CrystallineAmorphous

ITO

a

c

b

d

GST

Mirror

ITO t

t = 70 nmt = 120 nmt = 50 nm

0.8

0.6

0.4

0.2

1.0

t = 180 nm

650

600

550

500

450

400

350

700

750

Crystalline, simulated0.8

0.6

0.4

0.2

1.0

650

600

550

500

450

400

350

700

750

Amorphous, simulated

0.8

0.6

0.4

0.2

1.0

650

600

550

500

450

400

350

700

750

Crystalline, measured0.8

Reflectivity (arb

itra

ry u

nits)

0.6

0.4

0.2

1.0

300

200

100

400

t = 50 nm

amorphous

t = 120 nm

amorphous

t = 120 nm

crystalline

t = 50 nm

crystalline

t = 180 nm

crystalline

t = 70 nm

crystalline

t = 70 nm

amorphous

t = 180 nm

amorphous

630

nm

600

nm

580

nm

495

nm

450

nm

750

nm

350

nm

0

–100

300

200

100

400

0

0

–100

300

200

100

400

–100

80 95655035205

80 95655035205

80 95655035205

650

600

550

500

450

400

350

700

750

Amorphous, measured

Wavelength (nm)

GST thickness (nm)

ΔR (%

)ΔR

(%)

ΔR (%

)

t = 70 nm

t = 150 nm

t = 180 nm

Figure 1 | Colour tunability using ultrathin PCM films. a, Schematicrepresentation of the thin-film material stack comprising ITO/GST/ITO. Thecrystallization of the few-nanometres-thick GST layer induces a colour changein the entire film, visible when incident white light is reflected back.b, Simulated (top, left) and measured (bottom left) reflectivity spectra of theamorphous samples together with the simulated (top, right) and measured(bottom, right) reflectivity spectra of the crystalline samples shown in d. In bothphases, the close match between the model and simulation is striking. c, Changein reflectivity simulated for visible wavelengths, for various thicknesses ofthe ITO spacer and with increasing thickness of the GST layers. Each graphrepresents the change in optical reflectivity (Rcrystal 2 Ramorphous)/Ramorphous 3 100 when the PCM embedded in the ITO/GST/ITO/Pt structureis switched from the amorphous to the crystalline phase. By tailoring the

thickness of the ITO spacer, the reflectivity of specific wavelengths is enhanced,and the remaining wavelengths are either decreased or weakly increased, whenwhite light is used for illumination. d, An example of four different filmswith similar layered structure (starting from the topmost layer) consisting of10 nm ITO/7 nm GST/t nm ITO/100 nm Pt with different thicknesses of t, theITO layer deposited between the GST and the reflective Pt layer. The valueof t determines the reflected colour. After the deposition process, one part ofeach sample is placed on a hot plate at 220 uC for a few minutes to inducecomplete crystallization of the GST layer. The corresponding colour changeis also visible, and is quite dramatic. For example, the white (t 5 120 nm)amorphous film turns light blue when crystallized. No postproduction colouris added to enhance contrast.

LETTER RESEARCH

1 0 J U L Y 2 0 1 4 | V O L 5 1 1 | N A T U R E | 2 0 7

Macmillan Publishers Limited. All rights reserved©2014

optimized, along with the thicknesses of surrounding dielectric layers,to enhance the optical contrast. However, the GST thicknesses used incommercial disks are relatively high (several tens of nanometres, typ-ically). In the case of the films shown in Fig. 1d, the thickness of the GSTlayer is just 7 nm (the very minimum we could sputter reliably using ourfacilities). Counter-intuitively, as we explain below, smaller thicknessesenhance contrast dramatically and, importantly, lead to lower power elec-trical switching. Figure 1c shows our computational study of the change(shown as a percentage) in optical reflectivity for various thicknesses ofthe ITO spacer and with increasing thickness of the phase-change layer.For a thin layer of GST (that is, below 20 nm) the range of wavelengthsthat show increased reflectivity upon crystallization varies depending onthe chosen thickness of the ITO spacer. This allows the perceived colourof the entire stack to be transformed simply by changing the phase of aGST layer a few nanometres thick. Interestingly, this crystallization-inducedvariation of reflectivity is exponentially amplified for thinner layers ofthe PCM. Also, by tailoring the thickness of the ITO spacer, the reflec-tivity of specific wavelengths is enhanced when white light is used as theillumination source. Figure 1c shows three examples with ITO thicknessesof 70 nm, 150 nm and 180 nm where the red, green and blue components,respectively, are enhanced with the remaining wavelengths either decreasedor weakly increased. The importance of the ITO spacer is further eluci-dated in section S7 of the Supplementary Information, where the absenceof interference (that is, no ITO spacer is inserted between the GST and thePt layer) causes the reflectivity of the film to increase only weakly in thecrystalline phase, mostly independently of the wavelength.

We now demonstrate how this thin-film stack could be used in futurePCM-based display applications. To demonstrate the working principleof a display that can be electrically switched, we pattern an image using a

nanometre-sized electrical probe (that is, the conductive tip of an atomicforce microscope, CAFM), which was previously shown to be capable ofswitching nanoscale crystalline areas18–20 in PCM films. Our techniqueuses a pixelated approach mimicking real pixels in actual devices (thefabrication of such a device is described below). The use of CAFM tocharacterize the nanoscale switching characteristics of PCM from themore-amorphous to more-crystalline phase has been studied previously,and is known to closely resemble a standard cross-bar phase-changememory cell; hence, we employ this technique to demonstrate our elec-trically enabled display technology18,21. More details regarding this tech-nique and how it relates to actual PCM-based devices (pixels) can befound in section S1 of the Supplementary Information. To demonstratethe high optical contrast, we start by electrically drawing the OxfordRadcliffe Camera building (the grayscale image of which is shown inFig. 2b) in Fig. 2c, d and e on similar films with ITO spacer thicknessesof 50 nm, 70 nm and 180 nm, respectively. All the pictures are clearlyvisible using a standard, non-polarized, optical microscope. No artificialcontrast or colour has been added in postproduction. The striking con-trast demonstrates the feasibility of such ultrathin PCM-based microdis-plays. To demonstrate the colour-rendering capability, a different patternrepresenting the University of Oxford logo (shown in Fig. 2f) is alsoelectrically drawn on a 20 nm ITO/7 nm GST/70 nm ITO/100 nm Ptstack. The resulting optical image is shown in Fig. 2g, which closelymatches the original logo, representing the ‘Oxford blue’ in Fig. 2f (notethe striking contrast between the blue and white regions).

We then verify that a pixelated array, as would be necessary in a displaywhere each pixel can be randomly accessed and manipulated, is approxi-mated very closely by our CAFM approach. To perform this verification,we fabricate an array of 300 nm 3 300 nm pixels with 200 nm pitch on a

a

c d e

100

b

f g h

Top

electrode

GST

Bottom

electrode

Mirror

Figure 2 | Reflective display films. a, Schematicillustration of the electrically induced colourchanges in phase-change-based electronic displays.b, The original (greyscale) picture of the OxfordRadcliffe Camera used as a pattern. c, Electricallyconstructed image on a continuous ITO/GST/ITO/Pt stack with a 50-nm-thick ITO spacer; the whiteregions are regions where the phase is amorphous.No artificial contrast has been added inpostproduction. The pattern is once againelectrically reproduced on various stacks with ITOspacer thickness of 70 nm (d) and 180 nm (e). Scalebars are 10mm, which show the ultimate resolutionsuch displays are capable of achieving. f, Theoriginal image of the University of Oxford logo isused as a different pattern. g, Electricallyconstructed image on a ITO/GST/ITO/Pt stackwith 70-nm-thick ITO spacer—the blue regions areregions where the phase is transformed. h, To verifythe effectiveness of our blanket film approach weelectrically transformed an array of lithographicallydefined 300 nm 3 300 nm pixels (200 nm pitch) ona standard vertical stack with a 50-nm-thick ITOspacer; the optical contrast is strikingly evident,experimentally confirming that the pixelated arrayis well approximated by a CAFM-generated one.

RESEARCH LETTER

2 0 8 | N A T U R E | V O L 5 1 1 | 1 0 J U L Y 2 0 1 4

Macmillan Publishers Limited. All rights reserved©2014

standard vertical stack with a 50-nm-thick ITO spacer. The active volumeof each phase-change cell (that is, the GST layer) is restricted to a single pixelprecisely addressable by the nanoscale tip. These pixels were electricallyswitched using a nanoscale conductive tip. Figure 2h shows the resultingoptical image, where the optical contrast is once again striking. Further ex-periments using electrically switched single pixels, together with the result-ing high optical contrast, are shown in section S1 of the SupplementaryInformation.

We now move on to transparent substrates and electrically modulatetheir optical transmission. Future semi-transparent displays will findapplications in emerging technologies such as contact-lens-type displays22,‘smart’ eye glasses, windshield displays, for modulating light transmis-sion through windows23 or even for synthetic retina devices24. Such lightmodulation, achievable with resolution beyond the diffraction limit, isalso important for future ultrafast phase modulators, where a white lightbehind the semi-transparent display is selectively diffracted to produceinterference. For such a configuration the reflective platinum under-layer is not deposited and the entire structure (ITO/GST/ITO) is sput-tered directly on top of a transparent substrate, quartz in our case. As inthe previous case of reflective displays, the electrically induced crystal-lization (and re-amorphization) can be used as a means of creatingoptical contrast.

Figure 3a shows a schematic of the principle behind semi-transparent,phase-change-based displays. The calculated change in transmission(shown as a percentage) between the amorphous and crystalline phaseis shown in Fig. 3b. According to the model, a GST thickness of around75 nm to 85 nm gives the highest change in transmission for most wave-lengths. However, owing to the intrinsic absorption of the GST layer, the

opacity of the films quickly drops after a few nanometres (see section S6of the Supplementary Information). We therefore choose a 7 nm GSTlayer sandwiched between a 20-nm-thick ITO and a 40-nm-thick ITO asthe best compromise between contrast and transparency. We also con-firm the calculated change in absorption and transmission character-istics of such a film (shown in Fig. 3c) by measuring and comparing itstransmission (or absorption) spectra upon crystallization.

We use the CAFM (similar to the experiments on reflective substrates)to render pixelated images on these substrates. Figure 3e shows an exampleof an electrically constructed image on such a substrate. To illustrate thetransmissivity of the substrate, a square green pad is placed underneaththe rendered image; the transmissivity modulation is visible in Fig. 3dand 3e (the logo of Wolfson College, University of Oxford) and the contrastand transparency of the electrically drawn image is clear. This demon-strates the principle of using such phase-change optoelectronic frame-works for transmissive micro- and nanodisplay applications, from electronicdisplays on windows, to backlit displays and as future wavelength-tuneablewindows.

Finally, given the extremely low thickness of our films, we find thatthey are amenable to flexible films (and therefore flexible displays). Thedevelopment of a low-power, ultrathin flexible display has been a long-term goal worldwide, and the use of ITO could enhance the possibilityof using them as touch-sensitive displays. Figure 4a shows a picture ofa 20 nm ITO/7 nm GST/50 nm ITO/100 nm Pt multi-layered film onbiaxially oriented polyethylene terephthalate (boPET). The flexibility ofthese films with the sputtered layers is demonstrated in Fig. 4a, c and d.The colour variation in these flexible reflective displays is entirely sim-ilar to that on rigid substrates (Fig. 2). Again, we render images using a

e

b

GST thickness (nm)

ΔT (%

)

a

ITO

ITO

GST

d

c ΔT

(%

)/ΔA

b (%

)

Wavelength (nm)

Transmission

Absorption

40

20

0

–20

–40

–60

–80

–1005 20 35 50 65 80 95

750 nm

630 nm

600 nm

580 nm

495 nm

450 nm

350 nm

120

100

80

60

40

20

0

–20

–40

–60

50

400

450

500

550

600

650

700

750

Crystalline

Amorphous

Figure 3 | Semi-transparent display type films. a, Schematic diagramof the principle behind a semi-transparent device using a ITO/GST/ITOstructure on quartz. b, Computed change in optical transmission

DT (%) 5Tcrystal{Tamorphous

Tamorphous|100

� �, at different wavelengths, when the

GST layer is switched between the amorphous and crystalline phases.c, Measured change, upon crystallization, in absorption

DAb (%) 5Abcrystal{Abamorphous

Abamorphous|100

� �and transmission spectra of a

20 nm ITO/7 nm GST/40 nm ITO film sputtered on quartz. d, A picture of thefilm measured in c. e, Microscope picture of an electrically constructed imageon the film, demonstrating the principle of semi-transparent PCM-basedelectronic paper displays. The green sample pad visible underneath the imagedemonstrates the semi-transparent nature of the thin-film layers and enhancesclarity (because of the finite thickness of our transparent substrate, it is not infocus). The scale bar is 50mm, highlighting the high resolution that is possible.

LETTER RESEARCH

1 0 J U L Y 2 0 1 4 | V O L 5 1 1 | N A T U R E | 2 0 9

Macmillan Publishers Limited. All rights reserved©2014

nanoscale conductive probe, and the resulting image is shown in Fig. 4b.Although the quality of the boPET substrate is much inferior to that ofsemiconductor-grade SiO2 or quartz, the electrically written high-resolutionimage of the Oxford Radcliffe Camera is readily visible. Another exampleof a similar film with a 180-nm-thick ITO spacer is shown in Fig. 4c, wherethe colour is now closer to yellow, which shows the range of possiblecolours. Thus the use of such flexible films in semi-transparent modesis also possible. It is also remarkable that there is no visible colour var-iation when the films are bent, suggesting that such thin films also canbe used as wide-viewing-angle displays. A semi-transparent, flexible filmis shown in Fig. 4d, completing the demonstration for reflective and semi-transparent type displays on rigid, flexible and transparent substrates.

In all the above examples we use a CAFM to electrically switch theGST sandwiched between two ITO layers. Because no device using ITOelectrodes has yet been demonstrated, we show that such a device (asingle pixel) can be made and switched. Using a combination of electronbeam lithography, reactive ion etching and sputtering deposition, wefabricate vertical, crossbar-like ITO/GST/ITO devices to test their elec-trical switching characteristics. We fabricate these devices on 300-nm-thick SiO2 on silicon substrates (see section S3 of the SupplementaryInformation for fabrication details, and section S1 for the use of ITO asactive electrode). An optical image of a few crossbar devices is shown inFig. 5a with a false-colour magnified scanning electron microscopy imageof one such device in the inset. This particular device has an active area(defined by the area of overlapping crossbars) of 300 nm 3 300 nm.We note that switching in phase-change memories is often restricted tonanometric regions of the cell25. The use of nanoscale devices (pixels) istherefore preferable, since full crystallization and re-amorphization ofthe pixel is required to enhance these novel optoelectronic functional-ities. The electrical switching characteristics of such devices have beentested for various test cells with an example shown in Fig. 5b and c. Thisparticular cell shows a threshold voltage26 of 2.2 V with a 3503 increasein conductance between the amorphous and crystalline phases. Figure 5bshows a collection of direct-current (d.c.) current–voltage curves used totransform the device to the high-conductance (low resistance or crystal-line phase) state (referred to in PCM devices as the SET state). 100-ns

pulses with a fall-time of 5 ns and amplitude of 5 V are used to transformthe device to its initial low conductance (high resistance or amorphous)state (referred to as the RESET state in PCM devices). A collection of SET/RESET iterations is shown in Fig. 5c demonstrating its repeatability (thecycle endurance of our devices will require further engineering to create acommercial system, but we note that commercial-type phase-change mem-ories with endurances of greater than 108 cycles have been demonstrated27).It is clear that our device is a single pixel (of very high resolution), and itis reasonable to expect that such a pixel can be replicated for displayapplications.

We now address the energy requirements for such a display using theresults obtained from the single pixel device. In phase-change memoriesthe highest energy consumption is in the re-amorphization of a cell wheresufficient energy is required to melt the PCM4. We estimate the energyrequired to re-amorphize a single ITO/GST/ITO cell to be 15 pJ, on thebasis of our experimental results. A 300-cm2 display with 200-nm cell sizeand a 50-nm gap between each cell would thus require 7.2 J to completelyre-amorphize, similar to electrochromic devices28. The development ofeven thinner PCM would enable the energy consumption of a device(pixel) to be drastically reduced by ten to a hundred times13, which wouldalso increase the optical contrast, as already discussed. This would enablethe use of phase-change-based electronic paper also as dynamic displayswith few constraints on power requirements and ultrahigh (in the mega-hertz range) switching speed. Crucially for power consumption, no energy isrequired after a switching event, giving net zero-power consumption in staticmode, a further improvement over existing colour back-lit microdisplays.

Such cells can thus be fabricated, are reliable and can be switched re-versibly with low power. Many challenges remain, especially in understand-ing how larger pixels can be switched most effectively and yet displayoptical contrast, as well as in arraying and subdividing pixels effectivelyto create any arbitrary colour and greyscale. The use of growth-dominatedPCMs may prove to be a very useful approach in this respect, although

a b

c d

Figure 4 | Flexible display films in both reflective and semi-transparentmode. a, Reflective device on boPET, showing flexibility as well as the wide-viewing-angle colour consistency of such films (that is, when the film is bent,the colour in the bent region is not dramatically different). b, An electricallydefined image is patterned directly on top of the flexible substrate using aCAFM. c, Another example of a reflective stack. d, A semi-transparent typestructure sputtered directly on a flexible boPET substrate.

a b

Voltage (V)

SET / RESET number

Resis

tance (Ω

)

c

0.5 1 0 1.5 2 2.5 3 0

5

10

15

20

25

20

40

60

80 10

0 12

0 14

0

106

107

108

109

Curr

ent

(μA

)

Figure 5 | Demonstration of a single pixel. a, Optical image of a few ITO/GST/ITO crossbar-like devices, with a false-colour scanning electronmicrograph of one single device in the inset. The scale bar is 100mm. b, Acollection of various current–voltage curves demonstrating the amorphous-to-crystalline switching behaviour of several ITO/GST/ITO cells. These devicesdisplay a threshold voltage of approximately 2.2 V with a decrease in resistanceof about 3503 when switched from the amorphous to the crystalline phase. Themaximum current was limited to 20mA. c, SET/RESET cycling of the samedevice. Direct-current sweeps (similar to those in b above) are used to switchthe device from the RESET (high resistance) state to the SET (low resistance)state. Several pulses of 5 V in amplitude with 100 ns width, 20 ns rising and 5 nstrailing edges are used to switch the device back from the SET state to theRESET state.

RESEARCH LETTER

2 1 0 | N A T U R E | V O L 5 1 1 | 1 0 J U L Y 2 0 1 4

Macmillan Publishers Limited. All rights reserved©2014

their optoelectronic properties also remain unexplored and could bethe subject of further research in this field.

In conclusion, we have demonstrated a unique optoelectronic frame-work based on low-dimensional PCM films. The integration of a PCM asan active switching layer in displays paves the way for a new class of‘smart’ devices that are both electrically and optically active. Our resultsrepresent the first applications of the electro-optical properties of suchmaterials. Very thin layers of PCM can display very high optical contrastupon switching between the amorphous and crystalline phases whenlayered in appropriately designed thin film structures. Because this canbe done electrically, it allows for the fabrication of large arrays to createdisplays with ultrahigh resolution. We demonstrate such displays on re-flective and transparent substrates both on rigid and flexible surfaces;results that promise truly foldable ultralight displays that consume nopower in static mode. We then further demonstrate the fabrication ofone single pixel of a phase-change cell with transparent electrodes. BecausePCMs are also capable of storing data and performing arithmetic and logic,such devices could potentially mimic the functionality of photoreceptorcells in the human eye, opening up new fields of research in human–machine interaction, multi-functional glasses, contact lenses and syn-thetic retina devices.

METHODS SUMMARYFilm deposition. Films are sputtered directly on thermally grown SiO2 wafers (IDBTechnology, UK), boPET (Mylar) or quartz. Substrates were cleaned for 10 min inacetone under ultrasonic agitation, rinsed in isopropanol and dried with pressur-ized nitrogen. Where a reflector was needed (reflective devices), 100-nm-thick Pt wasdeposited using a Nordiko sputtering system: starting pressure 4.73 1027 torr, workingpressure 1.6 mtorr, 30 sccm (standard cubic centimetres per minute) Ar, 400 W d.c.9 r.p.m. The samples were then moved to a second sputtering system for ITO andGST deposition. ITO is sputtered from a solid target (Testbourne, UK) using 25 Wdirect current, 1.1 mtorr working pressure, 4 3 1026 torr starting pressure, 100 uCsample temperature at a rate of 2 nm min21. Substrate heating is found to greatlyreduce final surface roughness, increase optical transparency and electrical con-ductivity. Ge2Sb2Te5 is sputtered from a solid target (Super Conductor Materials,USA) at 25 W d.c., 1.1 mtorr working pressure, 4 3 1026 torr starting pressure anda rate of 8.5 nm min21, immediately following the ITO deposition in vacuum.Optical characterization. Films were measured using a Lambda 1050 spectrometer(Perkin Elmer, USA) in reflection, absorption and transmission using wavelengthsbetween 350 nm and 750 nm. Reflectivity measurements were calibrated against acommercial aluminium standard.Transfer matrix method. We employ a transfer matrix method described by Heavens29

and successfully applied to various systems17,30. More information can be found insection S5 of the Supplementary Information.Nanoscale patterning of continuous films and pixels. Patterns are generated byconverting a selected greyscale image into a spatially resolved voltage pattern. AnAsylum Research MFP-3D atomic force microscope equipped with an ORCA access-ory (modified to sustain our higher currents) and a conductive tip (DDESP, Bruker) isused as a nanoscale tip to electrically render several images on both films and pixels asshown in Fig. 2a.

Received 15 January; accepted 13 May 2014.

1. Wuttig, M. Phase-change materials—towards a universal memory? Nature Mater.4, 265–266 (2005).

2. Lankhorst, M. H. R., Ketelaars, B. W. S. M. M. & Wolters, R. A. M. Low-cost andnanoscale non-volatile memory concept for future silicon chips. Nature Mater. 4,347–352 (2005).

3. Chen , Y. C.et al. Ultra-thinphase-change bridge memorydevice using GeSb.2006International Electron Devices Meeting (IEDM 2006) http://dx.doi.org/10.1109/IEDM.2006.346910 (2006).

4. Raoux, S. et al. Phase-change random access memory: a scalable technology. IBMJ. Res. Develop. 52, 465–479 (2008).

5. Caldwell, M. A., Jeyasingh, R. G. D., Wong, H. S. P. & Milliron, D. J. Nanoscale phasechange memory materials. Nanoscale 4, 4382–4392 (2012).

6. Ohno, E., Yamada, N., Kurumizawa, T., Kimura, K. & Takao, M. Tegesnau alloys forphase-change type optical disk memories. Jpn. J. Appl. Phys. 28, 1235–1240(1989).

7. Afonso, C. N., Solis, J., Catalina, F. & Kalpouzos, C. Ultrafast reversible phase-change in GeSb films for erasable optical storage. Appl. Phys. Lett. 60, 3123–3125(1992).

8. Wuttig, M. & Yamada, N. Phase-change materials for rewriteable data storage.Nature Mater. 6, 824–832 (2007).

9. Kats, M. A., Blanchard, R., Genevet, P. & Capasso, F. Nanometre optical coatingsbased on strong interference effects in highly absorbing media. Nature Mater. 12,20–24 (2013).

10. Oosthoek, J. L. M. et al. Evolution of cell resistance, threshold voltage andcrystallization temperature during cycling of line-cell phase-change randomaccess memory. J. Appl. Phys. 110, 024505 (2011).

11. Loke, D. et al. Breaking the speed limits of phase-change memory. Science 336,1566–1569 (2012).

12. Wang, W. J. et al. Engineering grains of Ge2Sb2Te5 for realizing fast-speed, low-power, and low-drift phase-change memories with further multilevel capabilities.2012 International Electron Devices Meeting (IEDM 2012) http://dx.doi.org/10.1109/IEDM.2012.6479143 (2012).

13. Xiong, F., Liao, A. D., Estrada, D. & Pop, E. Low-power switching of phase-changematerials with carbon nanotube electrodes. Science 332, 568–570 (2011).

14. Lee, S. H., Jung, Y. & Agarwal, R. Highly scalable non-volatile and ultra-lowpowerphase-change nanowire memory. Nature Nanotechnol. 2, 626–630 (2007).

15. Yamada, N. & Matsunaga, T. Structure of laser-crystallized Ge2Sb21xTe5 sputteredthin films for use in optical memory. J. Appl. Phys. 88, 7020–7028 (2000).

16. Collings, N., Davey, T., Christmas, J., Chu, D. P.& Crossland, B. The applications andtechnology of phase-only liquid crystal on silicon devices. J. Displ. Technol. 7,112–119 (2011).

17. Burkhard, G. F., Hoke, E. T. & McGehee, M. D. Accounting for interference,scattering, andelectrodeabsorption tomake accurate internalquantumefficiencymeasurements in organic and other thin solar cells. Adv. Mater. 22, 3293–3297(2010).

18. Bichet, O., Wright, C. D., Samson, Y. & Gidon, S. Local characterization andtransformation of phase-change media by scanning thermal probes. J. Appl. Phys.95, 2360–2364 (2004).

19. Satoh, H., Sugawara, K. & Tanaka, K. Nanoscale phase changes in crystallineGe2Sb2Te5 films using scanning probe microscopes. J. Appl. Phys. 99, 024306(2006).

20. Hamann, H. F., O’Boyle, M., Martin, Y. C., Rooks, M. & Wickramasinghe, K. Ultra-high-density phase-change storage and memory. Nature Mater. 5, 383–387(2006).

21. Bhaskaran, H., Sebastian, A., Pauza, A., Pozidis, H. & Despont, M. Nanoscale phasetransformation in Ge2Sb2Te5 using encapsulated scanning probes and retractionforce microscopy. Rev. Sci. Instrum. 80, 083701 (2009).

22. Lingley, A. R. et al. A single-pixel wireless contact lens display. J. Micromech.Microeng. 21, 125014 (2011).

23. Llordes, A.,Garcia, G., Gazquez, J.&Milliron,D. J. Tunablenear-infraredandvisible-light transmittance in nanocrystal-in-glass composites. Nature 500, 323–326(2013).

24. Zrenner, E. Fighting blindness with microelectronics. Sci. Transl. Med. 5, 210ps216(2013).

25. Nam, S. W. et al. Electrical wind force-driven and dislocation-templatedamorphization in phase-change nanowires. Science 336, 1561–1566 (2012).

26. Redaelli, A., Pirovano, A., Benvenuti, A. & Lacaita, A. L. Threshold switching andphase transition numerical models for phase change memory simulations. J. Appl.Phys. 103, 111101 (2008).

27. Bez, R., Cappelletti, P., Servalli, G. & Pirovano, A. Phase change memories havetaken the field. 5th IEEE International Memory Workshop 13–16, http://dx.doi.org/10.1109/IMW.2013.6582084 (2013).

28. Heikenfeld, J., Drzaic, P., Yeo, J. S. & Koch, T. A critical review of the present andfuture prospects for electronic paper. J. Soc. Inf. Displ. 19, 129–156 (2011).

29. Heavens, O. Optical Properties of Thin Solid Films Ch. 4, 46–95 (Dover, 1991).30. Pettersson, L. A. A., Roman, L. S. & Inganas, O. Modeling photocurrent action

spectra of photovoltaic devices based on organic thin films. J. Appl. Phys. 86,487–496 (1999).

Supplementary Information is available in the online version of the paper.

Acknowledgements We thank R. Taylor for scientific discussions related to opticalspectroscopy measurements. We are grateful to M. Riede for discussions on themodelling aspects of our study. This research was supported by EPSRC via grantnumbers EP/J018783/1, EP/J018694/1 and EP/J00541X/2 as well as the OUP JohnFell Fund.

Author Contributions All authors contributed substantially to this work. P.H. and H.B.conceived and designed the experiments. P.H. performed the experiments with inputfrom H. B and C.D.W. All authors analysed the data. The manuscript was written byP.H. and H.B. with input from C.D.W.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of the paper.Correspondence and requests for materials should be addressed toH.B. (harish.bhaskaran@materials.ox.ac.uk).

LETTER RESEARCH

1 0 J U L Y 2 0 1 4 | V O L 5 1 1 | N A T U R E | 2 1 1

Macmillan Publishers Limited. All rights reserved©2014